In this story ⌄
- The October 29 switch preserves existing conversations.
- A familiar task is a better personal check than a model leaderboard.
A saved prompt can be a tiny piece of hard-won peace. You finally get the summary short enough, the tone less stiff, the bullet points under control. Then the app changes the model underneath it.
That is the practical story behind Anthropic's next switch. On October 29, Haiku 5.5 replaces Haiku 4.5 in the model selector for Free, Pro and Max users across Claude chat, Cowork and Claude Design, including desktop and mobile. The good news deserves its own sentence: your conversations are staying.
What happens to an old conversation
According to Anthropic’s change notice, opening an existing Haiku 4.5 conversation after the switch does not erase its history. Continuing it produces a response from Haiku 5.5. You can select another available model if you prefer.
Claude Code, the API, and Team and Enterprise are excluded. Until October 29, 4.5 remains under More models for affected users. This is a limited change, not a global retirement.
A faster model still has to do your job
Haiku 5.5 arrived on October 7. Anthropic pitches it at quick, repeated work such as summaries and time-sensitive tasks, and gives this Haiku generation an adjustable effort setting. Its speed claims come from the company; we have not independently benchmarked them.
The appealing part is easy to understand. A quick assistant fits the small jobs scattered through a day: cleaning up a note, shortening a message, finding the action buried in a paragraph. You should not need a ceremony to do any of those things.
But “quick” includes the time spent fixing the result. An instant summary that quietly loses the one exception you cared about has made the job longer.
Give it a familiar piece of work
Before the switch, save one short, non-sensitive input and an answer you consider good. Keep the criteria boring and observable: preserve three dates, retain a caveat, avoid adding a conclusion you never wrote. Then try that same task with the replacement.
This is a personal spot-check, not a scientific model ranking. It can still tell you something a leaderboard cannot: whether the change helps with the thing you actually open Claude to do.
If the answer feels slow or overworked, inspect the model and effort controls beside the send button. Anthropic says higher effort takes more time and tokens. Thinking and effort are separate controls; Haiku 5.5 does not offer a thinking-off switch in the Claude app.
We would start by changing one setting, then checking the same input again. Rewriting the entire prompt immediately makes it harder to tell what changed.
There is more on that habit in our guide to writing a better brief. The point is to protect a useful routine. A new model name is company news; whether your afternoon gets easier is the part that belongs to you.
Reporting checked October 23, 2026. This is an explanation of the announced change, not a hands-on model review.
Try this next.
A few small steps to turn the idea into something useful.
Go a little deeper.
Original sources, practical instructions, and places worth exploring.
Claude Haiku 4.5 is leaving the model selector
October 29 timing, affected plans, exceptions and what happens to existing conversations.
support.claude.comIntroducing Claude Haiku 5.5
The October 7 launch, intended tasks, effort control and qualified company performance claims.
anthropic.comChange the model, effort, and thinking settings
Where to change models and how effort affects speed and usage.
support.claude.comExternal links open in a new tab. References and further reading are selected for this story; inclusion is not an endorsement.
Something we should know? Send a tip or correction.
Your take.
A different angle. A good question. A hill you’ll happily die on. We’re listening.
Be the first to add a little signal.
